There can be no artificial intelligence without big technology.
At the same time, worrying issues are beginning to surface: for example, the technology does not all work the way it is prescribed, and can produce high error rates or discriminatory results. The opacity of AI means that we may not fully understand and control the technology.
The reason for this is a core attribute of AI that speaks volumes: it is fundamentally dependent on resources owned and controlled by only a handful of large tech companies.
Three key dimensions
The dominance of big Tech companies in AI is reflected in three key dimensions:
1. Data advantage: Companies with access to the broadest and deepest behavioral data are leading the way in developing consumer-grade AI products. This is reflected in tech companies extending this data advantage through acquisitions and mergers. Tech companies have accumulated enormous economic power, which has allowed them to embed themselves as core infrastructure in health, consumer goods, education, credit, and many other industries.
2. Computing power Advantage: AI is fundamentally a data-driven industry that relies heavily on computing power for training, tuning, and deploying models.
3, geopolitical advantages: At present, AI systems are not only commercial products, but also strategic economic and security assets of the country, AI companies have become a key lever in this geopolitical struggle.
Why "Big Tech"?
Making it easy for regulators to investigate big Tech's intrusive data surveillance, interference with user autonomy, monopoly, and discrimination.
Big tech has a series of knock-on effects on the broader ecosystem, inspiring and even forcing other companies to join in.
The tech industry as a whole and the government as a whole are increasingly dependent on big tech companies. The core business strategy of these companies is to make themselves the infrastructure, making themselves an indispensable link in many parts of the technology ecosystem.
Strategic focus
Against this background, the following strategic priorities are particularly important:
(1) Liability inversion: When harm occurs, companies should be allowed to prove that they did not cause harm, rather than the public and regulators stepping in to investigate, identify and find solutions after the harm has occurred.
2) Break down silos between policy areas, better address the impact of progress on one policy agenda on other policy agendas, and avoid companies taking advantage of inconsistencies between policies.
3. Determine whether policy approaches can effectively regulate industry behavior and adjust strategies in a timely manner to prevent technology companies from evading government regulation.
4. Move beyond a narrow focus on legislation and policy and embrace a broad-based theory of change.
Window of Action: The AI regulatory landscape
Focusing on the key needs of AI policy, the report elaborates on future AI regulatory strategies in terms of large-scale AI models, anti-competitive behavior, algorithmic accountability, data minimization, technology and financial capital, biometric monitoring, and international digital trade. Key points include:
1. Reducing tech companies' data advantage.
Data policy is AI policy, and taking steps to curb a company's data advantage is key to limiting the concentration of power in tech companies. The report therefore recommends:
Establish clear rules limiting the company's collection or generation of consumer data.
(2) In the process of formulating AI policies, the privacy law and competition law will be effectively linked to avoid companies taking advantage of the incompatibility of these rules to seek their own interests.
Optimize regulatory guidelines and enforcement measures to scrutinize companies' integration of data advantages so that law enforcement can intervene to stop misuse of data before harm occurs.
2. Reform competition enforcement methods to reduce technology concentration in the technology industry.
Curb the use of data by big tech companies to acquire other companies, and investigate and punish companies when they engage in anti-competitive behavior.
Advance the process of antitrust cases to provide antitrust enforcers with stronger tools to challenge abusive practices specific to the technology industry.
(3) Competitive analysis in the field of integrated technology policy. Identify scenarios in which Internet platform companies may use privacy measures to consolidate their advantages, and analyze how the excessive concentration of the cloud market has a knock-on effect on cybersecurity.
3, prevent the disorderly expansion of biometric monitoring in new fields such as smart cars.
Practice has proved that data protection laws are not in place to prevent the harm of biometric systems. In this environment, a blanket ban on companies collecting and using data in certain areas should be key to future policy interventions.
4. Prevent digital trade agreements from weakening national oversight of algorithmic accountability and competition policy.
Trade agreements include binding international rules that limit the scope for governments to regulate commercial companies. Because of the secrecy of the negotiations and their relative immunity to public political pressure, they have become the focus of intense lobbying by the tech industry for preferential treatment.
However, non-discrimination prohibition rules in trade agreements should not be seen as a tool to shield big Tech companies from foreign competition regulation.
In addition, the confidentiality of source code and algorithms in trade agreements should not be used as a tool to weaken the transparency of algorithms, and regulatory authorities should conduct more proactive and continuous monitoring of AI systems.
